U.S. Environmental Protection Agency ENERGY STAR Server Specification

Draft 4 Online MeetingMarch 16, 2009

Andrew FanaraEPA ENERGY STAR®

Product Development Team

Agenda

• Introductions/Announcements - 5 minutes

• Main Criteria Review - 15 minutes • Major changes in the Draft 4 – 1 hour 15

minutes • Questions and Answers - 25 minutes

2

Announcements

• Audio Bridge – Toll-free: +1 (800) 503-2899 – Toll: +1 (303) 248-0817 – Participant code: 8621266

• Please mute your line [*6] / unmute [*7] • Please hold all questions until end of presentation

– Submit verbally or through the Live Meeting Console(use “Q&A” on menubar)

• Slides will be posted to the ENERGY STAR Website following the meeting

3

Agenda

• Introductions/Announcements - 5 minutes

• Main Criteria Review - 15 minutes • Major changes in the Draft 4 – 1 hour 15

minutes • Questions and Answers - 25 minutes

4

ENERGY STAR for IT Equipment Goals

• Main goal is to transform the market for greater efficiency in IT Equipment– Expansive market coverage from basic low-end to

large complex systems. – Popularize advanced power management on both the

system and data center level. –Encourage the usage of more efficient components.– Standardize information reporting to understand

differences in energy consumption and operating expenses.

– Increase the manageability of systems in the data center. 5

Why ENERGY STAR for Servers?

• More organizations are recognizing theneed to confront energy & environmentalchallenges.

• Environmental attributes of products boughtor sold must be recognized, including theenergy used.

• ENERGY STAR helps partners share valuable information to confront these challenges, and in the process drives important business opportunities.

6

How Will ENERGY STAR Servers Provide Value to the End-User?

• Allow buyers to easily identify serverswhich focus on delivering performancewhile minimizing energy consumption. – For buyers looking to invest extra money in

performing work efficiently, not in achievingincremental performance gains.

• Greater transparency of information on the power use and capability of the system.

• Increased manageability through betteronboard technology.

7

Delivering Performance while Minimizing Energy Consumption (1)

• High Power Supply Efficiency for all systems– The power supply is the energy bottle neck of

the server – Established efficiency test procedures

available – Rewards systems with minimal power

conversion losses entering the server

8

Delivering Performance while Minimizing Energy Consumption (2)

• Minimize Idle power for products likely to be found idling in current market –For single and dual socket systems (1S & 2S)– Reduces energy wasted when servers are at

low utilization – Base levels plus adders approach allows for

expanded capability and performance – Rewards products with efficient components

and advanced power management

9

Delivering Performance while Minimizing Energy Consumption (3)

• Processor Power Management for systems more likely to be more heavily utilized – For systems with greater than two processor sockets

(3S & 4S) – Products more likely to take advantage of the benefits

of virtualization or to host heavily utilized applications

– Rewards products with advanced power management that scale power use with the amount of work required

– Systems still must test and report Idle power and make information available for end-user

10

Greater Transparency of Information

• ENERGY STAR Power and Performance Data Sheet will allow buyersand end-users to compareenergy use, thermalcharacteristics and performance of servers– Must be posted with product

specifications on mfg. Website

– A template for the data sheet will be provided on theEPA Web site

– Will reward mfgs who areproviding efficiency info ontheir products in an openand transparent way

11

Increased Manageability

• Data Measurement and Output Requirements for highly managed systems – Will require that ENERGY STAR servers have greater

ability to monitor and report their power consumption, input air temperature and processor utilization through standard networks

– Will give data center operators greater access to data on the operating state of their equipment

– Will reward systems with more advanced onboard technology to help enable greater efficiencies in the data center

12

Agenda

• Introductions/Announcements - 5 minutes

• Main Criteria Review - 15 minutes • Major changes in the Draft 4 – 1 hour

15 minutes • Questions and Answers - 25 minutes

13

Key Changes for Draft 4 Spec.• Removal of product labeling

–Required Web and product literature labeling • New PSU efficiency levels for smaller supplies at

low loading (≤ 500W) • New Idle levels for 1S and 2S Servers

– Idle Bins based on the installed processor count andlevel of manageability

–New adders for redundant PSU and I/O Devices

• Processor level power management for 4S servers –Must be enabled in hardware on shipment

14

Key Changes for Draft 4 Spec. (2)

• Qualification of Product Families –Based on common base components – One Power and Performance Data Sheet per Product

Family • Data Measurement and Reporting Requirements

–Exclusion for some 1S and 2S systems – Refined sampling and accuracy requirements to

reflect technologies currently on the market • New Idle test procedure based on ENERGY

STAR Computer V5.0 test procedure –Eliminates cost and complexity for new partners –Modified to include DC Powered Servers, Blades, etc.

15

16

Power Supply Efficiency and Power Factor Requirements

17

N = 8

Draft 3 Levels

Draft 4 Levels

Perc

ent E

ffici

ency

(%)

Single Output Power Supply Efficiency Req. for Small PSUs (≤ 500 W)

100%

95%

90%

85%

80%

75%

70%

65%

60%

0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50%

Load on Power Supply (%)

Single Output Power Supply Efficiency and Power Factor Req.

EfficiencyQualification Rate %

(%) pp % 8 2

30 9 % % % % 31 8

69 19

%

(%) pp 5 2 2

41 36

Requ irement 10% Load

Criteria 20% Load

Criteria 50% Load

Criteria

100Load

Criteria Total Unit

Count Pass All Loads Pass

Eff. Small Su lies 70.0 82.0% 89.0% 85.0% 25.0% Eff. Medium Sup plies 75.0% 85.0% 89.0% 85.0% 30.0% Eff. Large Supplies 80.0 88.0 92.0 88.0 25.8%

27.5% Power Factor Qualification Rate

Requ irement 10% Load

Criteria 20% Load

Criteria 50% Load

Criteria

100Load

Criteria Total Pass All Loads Pass

PF Small Su lies 0.6 0.80 0.90 0.90 100.0% PF Medium Supplies 0.65 0.80 0.90 0.90 16 13 81.3% PF LargeSupplies 0.80 0.90 0.90 0.90 23 21 91.3%

87.8%

Overall Qualification for units with both Efficiency and Power Factor data = 9/41 = 22.0%

18

Active Power Requirements

• Draft 4 continues to specify Idle levels forsingle and dual socket servers

• 3S and 4S systems now have processor power management requirement – These systems are more likely to be

virtualized and used in highly utilizedenvironments

–Highly configurable and upgradeable so Idle would be very complicated for 3S & 4S

– Manufacturers still need to test and reportIdle results to EPA 19

20

Idle Power Levels

• Managed Server must have: – Capability to operate with redundant power supplies – An installed dedicated management controller (e.g.,

service processor) • Base system includes: non-redundant PSUs,

Single Hard Drive, ≤ 4GB memory, up to dualport 1 Gbit onboard NIC

Idle Power Categories and Levels

6 65 9

100 8 150

161 42 %

Category Label Unit Count

Average Idle

Estar Level Pass Pass %

1P Systems 26 71.9 55 23.1% 1P Managed Systems 31 121.2 29.0% 2P Systems 30 153.8 26.7% 2P Managed Systems 74 238.7 19 25.7%

Sum Sum 26.1

• Categories for managed and standard servers for bothsingle and dual processor servers

• Good representation for each category • Increasing average Idle • Balanced qualification rate across categories

21

Idle Data and Base Idle Levels

22

0

100

200

300

400

500

600

700

800

900

()

1P 2P l l

Idle

Pow

er W

atts

1P Managed 2P Managed Base Id e Leve

23

Idle Power Adders

•New adders for redundant power supplies and I/O devices

•Hard drive adder changed to 8 Watts for all additional hard drives

•Memory adder threshold reduced to 4 GB

NEW

NEW

Idle Power Adders (2)

• Adders only included for key elements that determine the relative performance and capability of a general purpose Computer Server available in marketplace today

• Adders for RAID controllers and redundant fans suggested in Draft 3 comments – EPA believes products with these functionalities can

qualify with current base levels and adders • EPA not considering additional adders at this time

24

Redundant Power Supply Adder

• With new bins based on Managed/Standardservers, EPA has included an adder for theinstallation of redundant power supplies

• Allows for a more direct accounting of theadditional power required to operate redundant power

• Adder is for all power supplies added forredundancy (i.e. in addition to the minimumneeded to run the server)

25

Redundant PSU – Examples

Min. PSUs needed

PSUs installed Adder

1 1 No Adder 2 2 No Adder 1 2 1 x 20W = 20 W 2 3 1 x 20W = 20 W 2 4 2 x 20 W = 40 W

26

I/O Adder

• Base system includes up to 2 ports of onboard1GBit Ethernet –Anything greater gets the additional power allowance

• Adders only for devices ≥ 1 Gbit • 1Gbit and 10 Gbit Ethernet get additional power

per port • FibreChannel, InfiniBand get additional power

per device • Adders only applied for ports/devices active on

shipment • Draft 4 levels based on limited data

– EPA is in the process of collecting additional data to revise levels up or down. 27

I/O Adder – Examples

• Server with a 4 port onboard 1 Gbit Ethernetadapter would get a 2 x 2W adder (two 1 GbitAdders above the base)

• Server with a 2 port onboard 1 GBit Ethernetand a 4 port PCI 1 Gbit Ethernet Adapter wouldget a 4 x 2W adder (four 1 Gbit Adders abovethe base)

• Server with a 2 port onboard 1 GBit Ethernet; a2 port PCI 10 Gbit Ethernet adapter; and aFibreChannel device would get a 2 x 8W + 1 x5W (=21 W) adder

28

Idle Data and Base Idle Levels – 1PIdle Power and Levels for 1P Systems

Idle

Pow

er (W

atts

) Id

le P

ower

(Wat

ts)

250.0

200.0

150.0

100.0

50.0

0.0 Idle Power - Qualified

Idle Power - Unqualified Idle Limit

Idle Power and Levels for 1P Managed Systems

250.0

200.0

150.0

100.0

50.0

0.0

Idle Power - Qualified Idle Power - Unqualified Idle Limit

Idle Data and Base Idle Levels – 2PId

le P

ower

(Wat

ts)

Idle Power and Levels for 2P Systems

300.0

250.0

200.0

150.0

100.0

50.0

0.0

Idle Power - Qualified Idle Power - Unqualified Idle Limit

30

31

Idle Data and Base Idle Levels – 2PIdle Power and Levels for 2P Managed Systems

Idle

Pow

er (

Wat

ts)

Idle

Pow

er (W

atts

)

800.0

700.0

600.0

500.0

400.0

300.0

200.0

100.0

0.0

Idle Power - Qualified Idle Power - Unqualified Idle Limit

Idle Power and Levels for 2P Managed Systems (Continued)

800.0

700.0

600.0

500.0

400.0

300.0

200.0

100.0

0.0

Idle Power - Qualified Idle Power - Unqualified Idle Limit

PM Requirement for 3S and 4S

• All processors must be able to reduce their power use in times of low utilization by:– Reducing voltage and/or frequency through

Dynamic Voltage and Frequency Scaling (DVFS), OR

– Using processor or core reduced power states when a core or socket is not being used

• Capability must be enabled on the hardware level to qualify

32

Data Measurement and Reporting

• Required for: – 1S and 2S Managed Servers – All 3S and 4S Servers

• Exclusion for 1S and 2S systems that don’t meet the definition of Managed Servers– EPA intends to require this for ALL servers

under Tier 2

33

Data Measurement and Reporting (2)

• Refined sampling and accuracy requirements – Input power measurements: +/-10% accuracy

– Processor utilization measurements: +/-5% accuracy for CPU utilization less than 90%.

– Input air temperature measurements: +/- 3° C

– Hardware polling rates of at least one sample per second.

– Data must be averaged on a rolling basis over a specific time period. Default rolling average of 30 seconds is recommended.

34

Standard Information Reporting

• Looking for comments on the draft Power and Performance Data Sheet–Layout and format – Information required –Ability to implement

• Draft 4 Changes –SPECpower testing no longer required – Must test and report at least one benchmark score

(vendors choice) – Must test and report power at full load, along with

method of testing. • Questions

– Is the information useful to buyers? – Is any key information missing? 35

Idle Test Procedure

• Eliminated requirement to run SPECpower– Decreases cost and complexity and lowers

barrier to entry for new partners • Draft 4 contains draft Idle test procedure

– Based on Computer V5.0 Idle test procedure,modified for unique characteristics of servers

– Includes procedure for DC powered serversand Blade systems

36

Implementation - Families

• Manufacturers will now be able to qualify ProductFamilies –Must use identical model motherboard and processor– Relative numbers of components (PSUs, DIMMs, hard

drives, I/O Devices) may change, but components mustbe identical

–Submit data on Maximum and Minimum configurations–One data sheet per family – All units qualified with the family must still individually

meet all the ENERGY STAR requirements

37

Implementation - VARs

• Value Added Resellers (VARs)– Partners must work with VARS to ensure continued

compliance of altered configurations – Otherwise all indication of ENERGY STAR qualification

must be removed – VARs which re-brand products may sign on as partners

and qualify products • EPA is looking for feedback on how to best

accommodate various sales channels while maintaining the integrity of the ENERGY STARbrand

38

Implementation - Labeling

• Please note that manufacturers may not promote Servers as being ENERGY STAR qualified, or use the label, until the effective date of May 1, 2009. –New partners must sign a partnership agreement– Partners in other ENERGY STAR programs

(e.g., current computer partners) must still sign onto the Computer Server program

39

Tier 2

• Tier 2 Specification– Development to begin immediately after Tier 1

effective – Take effect 12 – 24 months after Tier 1– Focus on performance benchmark, regardless of

configuration (i.e. eliminate adders) – Raise all PSU targets (e.g., CSCI Gold for all

Single-Output PSUs) – Data Measurement and Reporting for all Servers– Energy-Efficient Ethernet (IEEE 802.3az)

40

Servers: Timeline• Draft 4 version released on February 20 • Comments on Draft 4 Due March 20 • Schedule moving forward

–Stakeholder Web Meeting: March 16 –Final Draft released: Early April –Web Meeting on Implementation: Mid April–Final Spec released/effective: by May 1, 2009 –Tier 2 plans initiated: March/April 2009

• Draft Tier 2 framework document to be distributed

• Tier 1 Specification will take effective immediatelyupon finalization –Will be in place 12 to 24 months

41

Agenda

• Introductions/Announcements - 5 minutes

• Main Criteria Review - 10 minutes • Major changes in the Draft 4 – 1 hour 15

minutes • Questions and Answers - 30 minutes

42

Contact / More Information

• Andrew Fanara, EPA ENERGY STAR–Fanara.andrew@epa.gov

• Arthur Howard, ICF International –ahoward@icfi.com

• Send Draft 4 Comments to Rebecca Duff, ICF International by March 20 – rduff@icfi.com

www.energystar.gov/productdevelopmentwww.energystar.gov/datacenters

43